融合用户关系表示和信息传播拓扑特征的信息传播预测  

Integrating User Relation Representations and Information Diffusion Topology Features for Information Propagation Prediction

作　　者：吴运兵[1] 高航 曾炜森 阴爱英 WU Yunbing;GAO Hang;ZENG Weisen;YIN Aiying(College of Computer and Data Science,Fuzhou University,Fuzhou 350108,China;Department of Computer Engineering,Fuzhou University Zhicheng College,Fuzhou 350002,China)

机构地区：[1]福州大学计算机与大数据学院,福州350108 [2]福州大学至诚学院计算机工程系,福州350002

出　　处：《计算机科学与探索》2024年第10期2738-2749,共12页Journal of Frontiers of Computer Science and Technology

基　　金：国家自然科学基金(62002064);福建省自然科学基金(2022J01116)。

摘　　要：信息传播预测旨在分析信息传播网络或社交媒体中信息扩散的规律,以理解和预测信息的传播过程。现有工作主要关注用户群体关系中的社交关系和动态影响关系,忽略了群体关系中用户间相似性关系和个体关系中内在因素对用户转发信息的影响。为此,提出了融合用户关系表示和信息传播拓扑特征的信息传播预测模型,从群体关系和个体关系两个层面分析用户受信息影响的可能性。在群体关系层面,构建用户共现次数图来学习用户相似性关系表示,随后融合了用户关系表示和信息传播拓扑特征来更全面地捕捉群体关系;在个体关系层面,融合了用户个体特征表示和影响因素向量,以捕捉不同的内在因素对刺激用户转发信息h的影响。实验结果表明,该模型在两个公开数据集上性能均有提升,在Memetracker数据集上MAP@k和its@k评价指标分别平均提升了6.54%和2.75%。Information propagation prediction aims to analyze the patterns of information spreading in social net-works and social media,thus understand and predict the information diffusion process.Recent researches have shown that information propagation is influenced by both group and individual relations.Existing works have mainly concentrated on group relations,including social relation and dynamic structural relation.They ignore a core group relation,i.e.,the co-occurrence user relation,and an important individual relation,i.e.,the user-preference relation,leading to incomplete modeling of the information propagation process.To address this issue,this paper comprehen-sively considers both group and individual relations,and proposes an information propagation prediction model that integrates user relation representations and information diffusion topology features.For the group relation,this paper con-structs a user co-occurrence graph to learn user similarity relation representations,which are then fused with infor-mation diffusion topology features to capture group relations.For the individual relation,this paper fuses user repre-sentation and influencing factors to capture internal factors on stimulating users to share information.Experimental results show that the performance of the proposed model on two public datasets is improved,and the MAP@k and hits@k evaluation indicators on the Memetracker dataset are improved by an average of 6.54%and 2.75%,respec-tively.

关 键 词：信息传播预测 拓扑结构 相似性 共同兴趣 

分 类 号：TP391[自动化与计算机技术—计算机应用技术]